Viable Quantifies Quality Customer Feedback with AI

There is an implicit assumption in most analytics solutions: the data analyzed and the insights derived are almost exclusively quantitative. That is, they refer to numerical data such as number of customers, sales etc.

But when it comes to customer feedback, perhaps the most important data is qualitative: text in resources such as feedback forms and surveys, tickets, chat and email messages. The problem with that data is that, while valuable, it takes domain experts and a lot of time to read through and classify. Or at least, that’s been the case until now.

This is the problem Profitable seeks to address. Viable, which touts itself as the only quality AI company to provide natural language customer feedback queries, today announced the close of a $5 million fundraiser primarily for growth, R&D and new hires.

Viable CEO and Co-Founder Dan Erickson has identified the company’s origins, its differentiators and its quality approach on customer feedback with VentureBeat.

From product-market fit to customer feedback via NLP

Erickson, an engineer by trade, co-founded Viable with his identical twin brother Jeff, a designer. Both have been in the tech industry for about 15 years and skipped college to get into business early on.

The two have held senior positions at several startups and their career paths have intertwined over the years, meeting in the middle in product management, as Erickson put it.

Ultimately, the Erickson brothers decided to start their own company, focusing on tackling the problem of product-market adaptation that had fascinated Dan over the years. That was the beginning of what was initially called ‘Viable Fit’.

The team has developed a product to help people run what is known as the “superhuman product-market fit process† The process centers around a survey, followed by some analysis to help founders and product owners find a roadmap for their products.

To make that work at scale, the Viable team has developed its own natural language processing (NLP) technology. They soon learned that this turned out to be the most valuable part of their entire approach.

Obtaining pull: push and pull

Viable gained traction with companies much larger than the traditional company finding a product-market fit. So the Erickson brothers decided to turn around and focus on their NLP models.

Viable stopped measuring product-market fit and started to focus on collecting customer feedback through various channels. Viable’s platform also offers a full analysis service that provides a written analysis in addition to the feedback. That recipe can be applied to areas such as product management, customer experience and marketing.

The analysis Viable offers can be accessed in two ways: push and pull. For push mode, a weekly report is sent out describing what happened in your customer feedback during the past week. The report includes things like top customer complaints, compliments, questions and requests. The size of the report varies from ten to a few hundred paragraphs.

When people read those reports, they usually have questions that they need to answer in order to act. Viable helps them with this by offering a question and answer system in natural language. Users can type a question about the data and Viable will provide an answer, all in plain English.

In addition, the company offers out-of-the-box integration with various sources such as Zendesk, Intercom, Delighted, iOS App Store, Play Store and Front. It features custom integrations via Zapier, as well as the ability to ingest data via .csv files. There are different subscription levels for the service, depending on the number of data points ingested.

Under the hood

It may sound simple and obvious, to the point where you have to wonder “how come no one else has done that before”. After all, Viable uses OpenAI’s GPT-3 under the hood, so in theory anyone – including the world’s Zendesks themselves – could have done it. The answer is twofold.

First off, Viable has a head start as they started in 2020 just when GPT-3 came out. As Erickson mentioned, they were one of the first to work with some of the capabilities of GPT-3 in a commercial environment. Second, part of Viable’s value proposition is the very fact that it integrates data from many different sources.

In fact, Viable is much more than a thin wrapper around GPT-3. The company uses many features of the OpenAI API, including embedding, as well as the actual GPT-3 complete engine. But Viable also has its own models that work with GPT-3, which have been trained and refined over the past two years.

The company also has its own data repository and its own recording pipeline. Whenever a new piece of content is created, it is brought in along with any metadata that may be available. From there it goes into a pipeline made up of several models that Viable has developed along with some GPT-3 functionality that will classify the piece of text.

The classification process determines whether the text is a complaint, a compliment, a request or a question. It also identifies various topics in the text and performs some sentiment analysis, emotion analysis, urgency analysis, and sound detection.

The platform is focused on text analysis and currently cannot directly connect to resources such as databases or spreadsheets. However, it can use what Erickson called “customer attributes” to slice and dice the data.

This could be job titles, locations, or even numerical answers to multiple choice questions such as “how many times a week do you use the product”. Users can then have the system perform tasks such as “generate a report for my product manager’s business customers in the Bay Area who use the product once or twice a week.”

Erickson said Viable has developed a completely unattended thematic analysis system based on GPT-3 embedding plus its own thematic analysis engine, which he labeled as state-of-the-art. That means the system doesn’t need to be provided with any context about what kind of things it’s looking for other than requests, questions, compliments, and complaints so that it can function in any domain.

Boundaries to avoid bias and toxic language

GPT-3 may be one of the most impressive feats of engineering and AI, but it’s not without its flaws. Two of the most well-known, which would make its use in a commercial setting problematic, are the generation of poisonous language and hallucination, ie, the generation of authoritative answers that are not based on fact. As Erickson shared, Viable has managed to get around these through custom training.

“We’ve developed thousands upon thousands of training examples for things like: What does it mean to summarize a theme? What does it mean to call a theme? How does all that work? And we’ve actually developed a fully refined version of GPT-3 that keeps it on track. So it kind of has a more limited language set that it uses. So it won’t do any of those curse words or anything like that,” Erickson said. “On the hallucinations side, we’ve done painstaking work building out that training dataset to make sure every single example we pull in only uses facts directly. from the feedback channeled into it. And that way it basically tells GPT-3 — Hey, I don’t want you to be creative here. I want you to just report the facts and that’s exactly how it works.”

Beyond GPT-3 and customer feedback

The above should be valuable free advice for anyone looking to build a business around something like GPT-3. Not only to get around the shortcomings, but also to add value to them. As Erickson said, the cost of using GPT-3 is baked into Viable’s price points, as are things like their other processing fees and healthy margin.

That must have worked for Viable’s investors. Streamlined Enterprises led the $5 million round for his interest in applied AI, with the participation of past investors Craft enterprises and Javelin Venture Partners† The round also included investments from Merus Capital, GTMFund, Stratminds, Tempo Ventures, Micheal Liou, Bill Butler and Samvit Ramadurgam. Viable’s total funding to date is now $9 million.

The company currently has about a dozen paying customers and a total workforce of nine employees. According to Erickson, the company has a few high-profile customers who are happy with the product, and Viable has taken the step to go beyond just customer feedback.

“We work for any kind of experience — whether it’s employee experience, partner experience, customer experience, it’s really about helping people analyze the qualitative nature of those experiences,” said Erickson.